32% at 27%: What the AI Sales Benchmark Actually Means for a Mortgage Branch

It's 4:50 on a Thursday, and a loan officer has fourteen open files, three underwriting conditions due by end of day, and a new pre-approval inquiry that came in an hour ago and is still sitting in the inbox. She'll get to it in the morning, she tells herself. By the time she does, the borrower has already talked to two other lenders and picked one.

That's not a discipline problem. It's math. One person can give full, competent attention to a handful of things at a time, and a working mortgage pipeline runs on files colliding for the same hour, every hour, all day, week after week.

The leak nobody's measuring

Most branches assume their win rate is set by how good their loan officers are at selling. It isn't, not entirely. It's set by how much of each incoming lead gets a timely, competent first touch before something else on the desk wins the loan officer's attention instead. And there's a lot competing for that attention: 40-60% of inbound leads arrive after business hours in the first place, landing in an inbox nobody's watching until the next business day.

That gap between "a good loan officer" and "a loan officer who actually has time to act like one" is the real leak. It doesn't show up on a scorecard or a commission statement. It shows up as pre-approvals that quietly go cold, week after week, without anyone being able to point to exactly what went wrong or where the file stalled.

Why it keeps happening

Mortgage branches hire and pay for closing skill, not for consistent intake handling. Nobody's actual job description is "answer everything within minutes and keep every file moving forward" — it's an expectation layered on top of a commission-driven role, so it happens when there's spare time and gets skipped when there isn't. There's rarely spare time, because the work that eats a producer's day — retyping an application by hand, chasing a missing pay stub, following up on a stale lead for the third time this month — is high-volume and repetitive but low-skill-per-task. None of it actually requires a loan officer's judgment. It just requires that someone do it, consistently, on a schedule, which is exactly the kind of work that quietly falls through the cracks in a busy branch.

The result is a branch that's staffed with genuinely skilled salespeople who spend a surprising share of their week doing something other than selling. Ask any loan officer to log an honest hour-by-hour breakdown of a Tuesday, and the selling conversations are usually a smaller slice of the day than either the loan officer or the branch manager expects. The rest goes to re-entering data that was already submitted once, tracking down a document that's already been requested twice, and returning calls to leads who reached out hours ago and have since moved on to whoever called them back first.

None of that is a knock on the loan officers themselves. It's a structural problem: the branch never built a system to separate the parts of the job that require a skilled salesperson's judgment from the parts that just require someone, or something, to follow through consistently.

A framework for getting more win rate out of the same team

None of this requires new headcount to fix. It requires separating the jobs currently stacked on one person and putting a structure around each one, starting today.

1. Split "intake" from "sale" into two distinct jobs, even in a small branch. Build a simple standing rule: every new inquiry gets an acknowledgment — even a short templated one — within a set number of minutes, before anyone decides whether it's a serious prospect. Assign that first response to whoever is free at that moment, not only to the loan officer the lead technically belongs to.

2. Use a short qualification script that anyone on the team can run immediately. A handful of questions — credit range, purchase timeline, property type, down payment situation — is enough to sort a workable lead from a tire-kicker. Don't let full underwriting-style qualification gate the first conversation; that's a later step, not a prerequisite for responding at all.

3. Put a decay date on every outstanding condition. If a borrower goes quiet on a missing document, someone follows up on a fixed schedule — three days, then seven, then a final check-in — instead of "whenever I remember to look at that file again." Put it on a shared calendar or task list, not in one person's head, so coverage doesn't depend on any single loan officer's memory.

4. Measure win rate against handling speed for one month, not against loan officer. Pull first-touch response time against deal outcome across your recent pipeline and look at the two side by side. Most branches have never actually run this comparison, and the shape of the data usually surprises them — it tends to make the leak visible for the first time in a way that individual performance reviews never do.

5. Protect your best producer's calendar for actual selling conversations. Every hour a loan officer spends on data entry or document chasing is an hour not spent talking to a borrower who's ready to move forward right now. The person doing the most producing should be doing the least admin work. In most branches today, it's closer to the opposite, and it's costing more than anyone's tracking.

Run these five changes for a single month before deciding whether they've worked. The pattern usually becomes obvious well before the month is out, and none of it requires hiring, firing, or restructuring the branch — just deciding, deliberately, who or what handles each part of a lead's first 24 hours instead of leaving it to whoever has a free minute.

What the data says about the gap

Across the framework Tykon has built and tested over several months of real deployment, automated sales handling reaches a 32% win rate using only 27% of the work a comparable human sales process requires — near parity with a trained human team's outcomes, but faster to execute and more consistent from lead to lead, with none of the attention-splitting problem described above.

That gap between win rate and effort is the number worth sitting with. It's roughly what's being left on a mortgage branch's table every time its best closers spend their week on intake and document chasing instead of on the phone with a borrower who's ready to sign. This is a general benchmark for the framework itself, not a claim about any single branch's results, but it's a useful gut check for where a branch's time is actually going relative to what it's producing.

Where this goes next

James, Tykon's AI sales agent, is built to be exactly the intake-and-chase layer described above: answering inbound loan inquiries within 60-90 seconds at any hour of the day or night, running the qualification script automatically and consistently, and keeping document follow-up on schedule so loan officers only get pulled into a file once it's genuinely ready to be sold rather than still being assembled.

Before any conversation about tools, though, there's a smaller exercise worth running on your own first: pull one week of inbound leads and time-stamp when each one actually got a real human response, not when someone meant to respond. That single number, by itself, usually tells you exactly where your branch's leak is sitting, and how much of the framework benchmark above your team could realistically close if that gap were closed first.